Background
Breast cancer, the second common cancer in females, is a complex heterogeneous disease, which can be divided into four major molecular subtypes [
1]. Of them, triple-negative breast cancer (TNBC), characterized by the loss of expression of estrogen receptors (ERs), progesterone receptors (PRs), and human epidermal growth factor receptor 2 (HER2), is more prone to metastasize to distant sites compared with other subtypes of breast cancer [
2]. Metastasis is responsible for over 90% of the cases of mortality in patients with breast cancer, and current effective therapeutic agent targeting metastasis is still lacking owing to the spatiotemporal intratumor heterogeneity during metastasis [
3]. Therefore, elucidation of the underlying mechanisms which contribute to metastasis is desperately needed to provide novel therapeutic strategies for patients with metastatic breast cancer.
Cancer metastasis is a complex and multi-step process [
4]. Accumulating evidence shows that epithelial-to-mesenchymal transition (EMT) is the pivotal step for breast cancer cells to metastasis, whereby epithelial cells gradually lose polarity and adhesion capacity but gain mesenchymal traits with down-regulation of the epithelial biomarker E-cadherin and up-regulation of mesenchymal markers vimentin [
5,
6]. Meanwhile, it is widely accepted that TGF-β signaling is a primary EMT inducer by activation of Smad complexes that translocate into the nucleus to regulate gene expression [
7], which is critical for breast cancer progression and heterogeneity [
8].
Recently, the roles of circular RNAs (circRNAs) in cancer have attracted much attention. CircRNAs, characterized by covalently closed loop structures without 5′-cup structure and 3′-polyadenylated tail, are highly conserved and stable [
9,
10]. With the advent of high-throughput sequencing, circRNAs are now known to be not simply by-products of splicing errors but rather the product of a new type of regulated alternative splicing [
11,
12]. CircRNAs are formed by back-splicing of exons or introns with gene-regulatory potency and both cis-elements (e.g., Alu elements) [
13,
14] and trans-acting factors (e.g., Quaking and Muscleblind) [
15,
16] participate in their biogenesis. Growing studies showed that circRNAs were involved in various cancer biological processes, including EMT and metastasis, via interaction with miRNA [
17,
18]. For instance, circRNA-MYLK promoted bladder carcinoma metastasis by inducing EMT via sponging miR-29a [
19]. CircSMARCA5 directly bound to miR-17-3p and miR-181b-5p to inhibit the metastasis of hepatocellular carcinoma [
20]. And our previous study also demonstrated that circHIPK3 contributed to colorectal cancer metastasis through interacting with miR-7 [
21]. Despite advancements in the study of circRNAs, the potential correlation between circRNAs and breast cancer metastasis is still unclear and remains to be further investigated.
In the present study, we characterized one novel circRNA originated from exons 5 to 8 of ANKS1B gene (hsa_circ_0007294, circANKS1B). Furthermore, the biogenesis, functions and mechanisms of circANKS1B in breast cancer were studied.
Methods
Patient population and clinical data
In total, 23 pairs of fresh frozen TNBC and adjacent normal tissues (cohort 1, 2), 165 formalin-fixed, paraffin-embedded (FFPE) breast cancer tissues and 40 normal tissues (cohort 3) were collected from Affiliated Nanjing First Hospital of Nanjing Medical University (Nanjing, China). Patients treated with any anti-tumour treatment before surgery were excluded. All histologic slides were independently identified by two pathologists. Of them, RNA-sequencing was performed in 3 pairs (cohort 1), 20 pairs (cohort 2) were used for circRNA validation and cohort 3 was used for quantification of circANKS1B in all breast cancer subtypes and normal tissues and analysis of the correlations between circANKS1B expression and breast cancer clinico-pathological parameters and outcome. The detail patient characteristics were described in Additional file
1: Table S1. Patient follow-up was performed in the outpatient department by phone or letter. Informed consent was obtained from each patient and the study was approved by the ethics committee of Nanjing First Hospital.
RNA sequencing, identification and quantification of human circRNAs
The total RNA was extracted from three pairs of fresh frozen TNBC and adjacent normal tissues by using TRIzol reagent (Invitrogen, CA, USA), followed by treatment with the RiboMinus Eukaryote Kit (Qiagen, Valencia, CA) to delete ribosomal RNA according to the manufacturer’s guidelines. Next, the processed RNAs were subjected to perform deep sequencing with an Illumina HiSeq 3000 (Illumina, San Diego, CA).
The RNA-seq FASTQ reads were first aligned to the human reference genome (GRCh37/hg19) by TopHat2 [
22]. The sequences that aligned contiguously and full length to the genomes were discarded. Then, the remaining reads were used to identify circRNAs [
14]. SRPBM (spliced reads per billion mapping) was applied to normalize the counts of reads mapping across an identified backsplice, and differential expression analysis was conducted based on the previous method [
23].
Cell culture
Human normal breast epithelial cell line (MCF10A) and breast cancer cell lines (MCF-7, T47D, SK-BR-3, MDA-MB-231, MDA-MB-468 and BT549) were all purchased from American Type Culture Collection (Manassas, VA, USA). The culture method of MCF-10A cell was described in the previous study [
24]. Other cell lines were cultured in DMEM or RPMI1640 media plus 10% fetal bovine serum (Gibco, Vienna, Austria) at 37 °C with 5% CO2. All cell lines were authenticated and tested for mycoplasma every 4 months using MycoAlert Mycoplasma Detection Kit (Lonza, Switzerland).
Antibodies and reagents
The antibodies we used are as follows: anti-E-cadherin (Abcam # ab40772), anti-Vimentin (Abcam # ab92547), anti-Fibronectin (Proteintech # 15613–1-AP), anti-AGO2 (Abcam # ab57113), anti-USF1 (Santa Cruz # sc-390,027), anti-RNA polymerase II (Santa Cruz # sc-47,701), anti-TGF-β1 (Abcam # ab92486), anti-ESRP1 (Abcam # ab107278), anti-p-Smad2 (Cell Signaling Technology # 8828), anti-p-Smad3 (Cell Signaling Technology # 9520), anti-Smad2/3 (Cell Signaling Technology # 8685), anti-GAPDH (Proteintech # 10494–1-AP) and anti-β-actin (Cell Signaling Technology # 4970). Actinomycin D and RNase R were purchased from Sigma-Aldrich (St Louis, MO, USA) and Epicentre Technologies (Madison, WI, USA), respectively. The TGF-β receptor type I/II inhibitor LY2109761 was obtained from Selleckchem Chemicals (Houston, TX, USA).
RNA extraction and qRT-PCR
Total RNA was isolated from tissue samples and cultured cells with TRIzol reagent (Invitrogen). RNA quantity was tested by a SmartSpec Plus spectrophotometer (Bio-Rad). The Hairpin-itTM MicroRNAs Quantitation PCR Kit (Gene-Pharma, Shanghai, China) was used to measure the expression of miRNA, and U6 was used as the internal control. To detect circRNA and mRNA, 1 μg of RNA was reverse transcribed to cDNA using the PrimeScript RT Reagent (Takara, Otsu, Japan) and then subjected to qPCR using the SYBR Premix Ex Taq™ (Takara). GAPDH was used as an internal control. The 2
−ΔΔCt method was applied to quantify gene expression. The primer sequences were shown in Additional file
1: Table S3.
Oligonucleotide transfection
siRNA, miRNA mimics and inhibitors were purified and synthesized by RiboBio (Guangzhou, China) or Gene-Pharma (Shanghai, China). Transfection was performed using Lipofectamine 2000 reagent (Invitrogen). The RNA sequences used are listed in Additional file
1: Table S3.
Plasmids construction and stable transfection
The circRNA-expressing vectors were constructed as described previously [
15,
25]. In brief, full-length of human circANKS1B along with 1.2 kb endogenous 5′-flanking intron and 0.8 kb 3′ -flanking intron was subcloned into the pCDH-CMV-GFP vector (Geenseed Biotech, Guangzhou, China). And the sequence of the 5′-flanking intron was copied and inversely inserted the downstream of 3′-flanking intron. Besides, the canonical splicing site (AG-GT) was reserved for correct splicing. For circANKS1B minigene reporters, the inversely inserted 5′-flanking intron was deleted. For SYT8 and Snail minigenes, genomic regions comprising three exons and two introns with or without ESRP1 binding site (GGT-rich) were synthesized and inserted into the pCDH-CMV-GFP vectors. For USF1, TGF-β1 and ESRP1-expressing vectors, the full-length ORF sequences of these three genes were respectively subcloned into the pLenti-CMV-GFP vectors (ABM, Richmond, BC, Canada). And two si-circANKS1B sequences were subcloned into pGLV3/H1/GFP/Puro vectors to construct sh-circANKS1B for animal studies. All constructs were verified by sequencing. Lentiviral particles carrying above-mentioned vectors were generated in HEK293T cells. Then, breast cancer cells were infected with lentivirus at a multiplicity of infection (MOI) of 30, followed by selection with 1–2 μg/mL puromycin.
Fluorescence in situ hybridization
The FISH assay was carried out using Fluorescent In Situ Hybridization Kit (Gene-Pharma, Shanghai, China) based on the manufacturer’s protocols. The hybridization was performed with Cy3-labeled circANKS1B and FAM-labeled miR-148a-3p or miR-152-3p probes (Gene-Pharma), followed by analysis using a confocal microscopy. The probe sequences were shown in Additional file
1: Table S3.
Wound healing and transwell assays
For wound healing assay, breast cancer cells were seeded into a 6-well plate and scraped using a sterile pipette tip. Images were obtained using an inverted microscope at 0 and 24 h, and then analyzed by Image J. For transwell migration and invasion assays, breast cancer cells were seeded into the upper chamber without (migration assay) or with (invasion assay) the matrigel (Corning, NY, USA). After 24 h of incubation, non-migrated or invaded cells were scraped off with a cotton swab and cells on the bottom of the chamber were fixed, stained, and counted.
Immunoblot analysis
Breast cancer cells were washed and then lysed in RIPA lysis buffer. After that, protein extracts were boiled for 5 min, separated on a 10% SDS-PAGE and transferred to a PVDF membrane (Millipore, Schwalbach, Germany). Subsequently, the membrane was incubated with corresponding primary antibody at 4 °C overnight. Next, the membrane was washed five times and incubated with secondary antibody, and bands were then visualized.
RNA immunoprecipitation
RNA immunoprecipitation (RIP) assay was performed using Magna RIP™ RNA-binding protein immunoprecipitation kit (Millipore) according to the manufacturer’s guidelines with minor modifications. Briefly, the magnetic beads were incubated with 5 μg anti-AGO2 or anti-ESRP1 antibodies for 30 min at room temperature to generate antibody-coated beads. Breast cancer cells (2 × 107) were lysed in 100 μl RIP lysis buffer and then diluted with 900 μl RIP immunoprecipitation buffer and incubated with the antibody-coated beads overnight at 4 °C. After that, beads were washed six times using RIP wash buffer. The immunoprecipitates were treated with Proteinase K at 55 °C for 30 min. And the isolated RNA was extracted using TRIzol regent (Invitrogen), followed by qRT-PCR.
Biotinylated RNA pull-down assay
The RNA pull-down assay was conducted as described previously [
21,
26]. Briefly, for the assay of pulling down miRNA by circRNA, breast cancer cells (1 × 10
7) were lysed and incubated with biotinylated
-circANKS1B probe that was pre-bound on C-1 magnetic beads (#65001, Life Technologies) at 4 °C overnight. Next, beads were eluted with rotation at 37 °C for 30 min. The bound RNAs were extracted for qRT-PCR. For the assay of pulling down circRNA by miRNA, breast cancer cells with circANKS1B overexpression were respectively transfected with biotinylated wild-type or mutant miR-148a-3p/ miR-152-3p mimics. 48 h later, the cells were collected and incubated with C-1 magnetic beads on the rotator at 4 °C for 3.5 h. And then washed five times and the bound RNAs were extracted for qRT-PCR. The probe sequences were described in Additional file
1: Table S3.
Luciferase reporter assay
The circANKS1B or USF1 3′ UTR sequences containing wild-type or mutant miR-148a/152-3p binding sites were synthesized and respectively inserted into pmirGLO luciferase reporters (Promega) between Sacl and Sall restriction sites, after which cotransfected with miR-148a/152-3p mimics or control mimics into breast cancer cells using Lipofectamine 2000. For the promoter of TGF-β1 luciferase reporter assay, the wild-type or mutant full-length TGF-β1 promoter construct and six truncation constructs were respectively inserted into pGL3-basic vectors (Promega) between Sacl and Xhol restriction sites, and then cotransfected with USF1 overexpression vector and pRL-TK into breast cancer cells by Lipofectamine 2000. After 48 h, the luciferase activities were tested by the dual-luciferase reporter assay kit (Promega).
Chromatin immunoprecipitation assay
The Chromatin immunoprecipitation (ChIP) assay was carried out as described previously [
21]. In brief, breast cancer cells were collected and sonicated to generate DNA fragments of 200–1000 bp and then incubated with anti-USF1, anti-RNA polymerase II (positive control) or anti-IgG antibody (negative control) overnight at 4 °C. Immunoprecipitated DNA was extracted and subjected to PCR analysis. The primer sequences were listed in Additional file
1: Table S3.
Immunohistochemistry
Immunohistochemistry (IHC) was performed as described previously [
27] with anti-USF1 and anti-ESRP1 antibody in the formalin-fixed, paraffin-embedded breast cancer tissue sections (
n = 165).
Animal studies
MCF-7 cells with circANKS1B overexpression, MDA-MB-231 cells with circANKS1B knockdown and their respective control vectors were respectively tail-vein (2 × 106 cells) injected into the female BALB/c nude mice (8 mice in each group). Six weeks later, mice were euthanized. The lungs were collected and metastatic nodules were counted after H&E staining. And the animal studies were approved by the Animal Care Committee of Nanjing Medical College (acceptance no.: SYXK20160006).
Analysis of public databases
The raw gene expression data in breast cancer (
n = 1109) were downloaded from The Cancer Genome Atlas (TCGA) database (
https://cancergenome.nih.gov/). Then, the expression values (counts) of USF1, TGF-β1 and ESRP1 were obtained by using R software. To evaluate the prognostic value of USF1 and ESRP1 in breast cancer patients, we analyzed the Kaplan-Meier plotter database (
http://kmplot.com/analysis/). The median expression values of USF1 and ESRP1 were set to the cutoff values of the overall and distant metastasis-free survival curves.
Statistical analysis
The differences between groups were determined by Student’s t-test or one-way ANOVA. Kaplan-Meier plot and Cox proportional hazards model were respectively applied to determine the patient survival and independent factors. And the correlations were measured by Spearman correlation coefficients. A two-sided p < 0.05 was considered statistically significant.
Discussion
In the present study, we identified a large amount of circRNAs by RNA-seq. And we then characterized one of the most differentially expressed circRNAs, circANKS1B, which was highly associated with breast cancer invasion and metastasis and poor prognosis. Functionally, circANKS1B promoted breast cancer cell invasion and metastasis without affecting cell proliferation and apoptosis. Mechanistically, circANKS1B could up-regulate USF1 expression by sponge activity of miR-148a-3p and miR-152-3p. Further, USF1 transcriptionally elevated ESRP1 and TGF-β1 expression through directly binding to their promoters, thereby activating TGF-β1 signaling to enhance EMT and metastasis. Besides, we found that ESRP1 increased circANKS1B production via interaction with its flanking introns. Thus, our findings identify a novel feedback loop that promotes breast cancer metastasis, which advance the understanding of molecular mechanism involved in the metastasis of breast cancer.
Accumulating evidence shows that circRNAs are abundant, stable and highly conserved in eukaryotes with gene-regulatory potency [
12,
17]. Here, we also identify a large number of circRNAs (69,815) and most of them are generated from precursor mRNAs by exon circularization. Amino acid sequence alignment shows that the similarity of human circANKS1B with homolog in
Mus musculus is 88% (data not shown), suggesting circANKS1B is a well-conserved gene. The abundance, stability and specific expression patterns of circRNAs allowing circRNAs to be the promising potential cancer biomarkers. Up to now, many circRNAs have been identified as diagnostic and prognostic biomarkers in human malignancy, such as colorectal cancer (CiRS-7 and circHIPK3) [
21,
34], gastric cancer (circPVT1 and hsa_circ_0000096) [
35,
36], hepatocellular carcinoma (circMTO1 and circSMARCA5) [
20,
37] and bladder cancer (circMYLK and circITCH) [
19,
38]. In this study, we found that breast cancer patients with higher circANKS1B expression displayed significantly worse overall survival, and high circANKS1B was an independent factor for poor outcome, as demonstrated by Cox proportional hazards model. These indicate that circANKS1B may be a promising prognostic biomarker in breast cancer.
It has been reported that circRNAs, like LncRNAs, exerted diverse biological functions by acting as miRNA sponges [
39‐
41]. Of note, due to the covalently closed structure of circRNA, it may maintain the miRNA-regulatory function for a longer period of time than LncRNA. Herein, using various assays, we found that circANKS1B promoted breast cancer invasion and metastasis, mainly through interaction with miR-148a-3p and miR-152-3p. The miR-148/152 family consists of miR-148a-3p, miR-148b-3p and miR-152-3p, is proposed to be potential metastasis suppressors in many cancers, including breast cancer [
42]. Mature miR-148/152 family shares similar seed sequence, which is a key region for regulating their targets [
30]. However, the RNA pull-down results showed that miR-148a-3p and miR-152-3p, but not miR-148b-3p, were abundantly pulled down by circANKS1B probe in both MCF-7 and MDA-MB-231 cells, revealing that miR-148b-3p might not be involved in circANKS1B-mediated metastasis-promoting process in breast cancer, this was also confirmed by the rescue experiment that inhibition of miR-148b-3p could not rescue the decreased migratory and invasive capabilities of breast cancer cells caused by silencing of circANKS1B (Additional file
1: Figure S12A-B). By a series of screenings and validations, we identified that USF1, the common target of miR-148a-3p and miR-152-3p, participated in circANKS1B-mediated pro-metastasis process in breast cancer. USF1 is a transcription factor and it can regulate the expression of different genes by binding to the E-box motifs (CANNTG) in their promoter regions [
31]. As the previous study reported that USF1 could bind to murine TGF-β1 promoter [
32], we then wonder whether this phenomenon also occurs on human TGF-β1 promoter. ChIP-qPCR and luciferase reporter assays clearly showed that USF1 could directly bind to human TGF-β1 promoter to transcriptionally up-regulate TGF-β1 expression. TGF-β1 is a well-known driver of EMT, which is critical for breast cancer metastasis [
7]. Therefore, these data show that circANKS1B, as miR-148a-3p and miR-152-3p sponge, relieves the miR-148a/152-3p-mediated inhibition of USF1 which subsequently transcriptionally elevates TGF-β1 to induce EMT, thus promoting breast cancer invasion and metastasis, supporting the notion that circANKS1B is capable of functioning as a miRNA sponge to modulate gene expression in breast cancer.
Recent studies showed that splicing factors played crucial roles in cell-type-specific circRNA formation [
15,
16]. Here, we identified a splicing factor, ESRP1, was involved in circANKS1B biogenesis in breast cancer. We found that ESRP1 promoted circANKS1B formation by interaction with “GGT-rich” motifs on upstream and downstream of the introns flanking circANKS1B-forming exons. Silencing of ESRP1 decreased circANKS1B expression by about 55%, suggesting that other splicing factors might participate in circANKS1B production. As shown in Additional file
1: Figure S10A, knockdown of ESRP2, a close family member of ESRP1 that recognizes the similar binding motif, partially reduced circANKS1B expression, and knockdown of QKI, a splicing factor was proposed to regulate over one-third of EMT-related circRNAs expression, could decreased circANKS1B expression by about 25%, implying that circRNAs biogenesis may be simultaneously controlled by multiple splicing factors. In addition, we found that ESRP1 disruption increased linear ANKS1B expression, this may be explained by the competition between circRNA and its linear isoform during pre-mRNA splicing [
16]. Importantly, mutation of the ESRP1-binding motifs on the flanking introns dramatically decreased circANKS1B formation, whereas exons that do not normally give rise to circRNAs could be capable of generating circRNAs by insertion of ESRP1-binding motifs into the flanking introns in breast cancer cells. And a recent study demonstrated that ESRP1 could also promote the generation of circBIRC6 in human embryonic stem cells [
25]. These indicate that ESRP1 may regulate multiple circRNAs and further studies will be warranted to explore the role of ESRP1-mediated circRNA biogenesis in other diseases. Moreover, we showed that ESRP1 was also a direct transcriptional target of USF1, providing evidence to further support the idea that the ESRP1-circANKS1B axis is a metastasis-associated regulatory pathway.
Although the metastasis-promoting effect of circANKS1B in breast cancer was illustrated in our study, owing to the limited breast cancer tissues we screened initially, we do not rule out this possibility that there may be other key dysregulated circRNAs which are also involved in breast cancer metastasis, as well as some other pathological processes such as proliferation or apoptosis. Therefore, the dysregulated circRNAs in breast cancer still need further elucidation.